[lx-0]

Closes #31.

What

A status line script can now append one compact ytstack segment for the directory it runs in.

V2 76% 138/180 · M021/S01/T02 · waiting: review of PR 12   (progress block)
M021/S01/T02 3/7                                           (no block: slice checkboxes)

• bin/ytstack-status: reads cwd from the status line JSON on stdin (else $PWD), prints the segment and exits 0, or prints nothing and exits 1 (no .ytstack/, nothing to show, malformed block). NO_COLOR / --no-color for plain text. Local refs only, no network, no gh.
• Optional progress block in the STATE.md frontmatter (label, done, total, waiting_on). Without it the segment falls back to the task checkboxes of the active slice plan.
• Two levels: plan progress is read from the newer of origin/<default> and <default> when the checkout is on another branch (a linked worktree carries a stale copy); the task pointer always comes from the checkout.
• yt_progress_parse in hooks/lib/git-ctx.sh is the only parser; yt_state_lint reports a malformed block as W7, so the segment and the lint cannot disagree.
• summarize-task and quick-task update the block when they advance STATE.
• docs/statusline.md: block reference, output forms, integration snippet, limits. Pointers in README, docs/concept.md, docs/codex.md, CHANGELOG [Unreleased].

Decisions

• The issue's open questions: exactly one track with done / total. Several named tracks and a done / open pair are not built (DECISIONS 2026-11-11).
• Locating the script: Claude Code documents no stable path to plugin files for a statusLine command, so the documented snippet globs the plugin cache and picks the highest installed version. A SessionStart-maintained link was dropped because that hook is read-only by contract.
• Planned as M021: M020 is taken by the Kimi adapter planning (feat: Kimi Code CLI host support — skills + lifecycle hooks via extra_skill_dirs and hook wrappers (working setup attached) #30), which is not committed yet. That planning also edits STATE.md and appends to DECISIONS.md, so expect a textual conflict there when both lines meet.

Verification

• bash tests/run-all.sh: 13 suites pass on macOS (bash 3.2), including the new tests/ytstack-status.test.sh (31 cases: outside a project, slice fallback, block, malformed block, input fallbacks, colour switch, linked worktree, detached HEAD, clone with a newer origin/main) and 16 new parser/lint cases in tests/git-ctx.test.sh.
• bin/ytstack-check: no failures, 27 warnings, the same count as main.
• python3 -m unittest discover -s tests -p 'codex*_test.py': 13 tests OK. claude plugin validate .: passed.
• The integration snippet was run against a fake plugin cache with versions 0.4.2 (no script), 0.9.0 and 0.10.0 and picked 0.10.0.
• Not run locally: tests/run-all.sh --linux (no Docker daemon). The ubuntu CI job on this PR is the Linux evidence.

Known limits

• waiting_on is cut by characters of the current locale; under LC_ALL=C a multi-byte character at the cut can be split (marked shortcut: in the script).
• One run takes about 90 ms on the development machine.
• The skill instructions for keeping the block current are prose; no eval case covers them yet.

🤖 Generated with Claude Code

---

Note

Low Risk
Additive CLI and frontmatter parsing with broad shell tests; shared git-ctx.sh change is limited to optional progress linting and does not alter hook write paths.

Overview
Adds a status line segment so any host can append one line of ytstack context: optional release-style progress frontmatter in STATE.md (label, done, total, waiting_on), or fallback to the active slice’s task checkbox counts plus the local M###/S##/T## pointer.

bin/ytstack-status reads cwd from status-line JSON (else $PWD), resolves the project via the existing git-ctx helpers, and prints a single formatted line (colour optional). On non-default branches and linked worktrees, plan progress comes from the newer of local/origin default-branch STATE.md via git show; the task pointer always comes from the checkout. Malformed blocks produce no output; yt_progress_parse in hooks/lib/git-ctx.sh is the sole parser, with yt_state_lint W7 so lint and the segment cannot disagree.

summarize-task and quick-task (0.1.1) document keeping the block in sync when STATE advances. docs/statusline.md covers the block, behaviour, and a plugin-cache glob snippet for Claude Code; README, CHANGELOG, concept.md, and codex.md point at it. Milestone M021 closes in .ytstack/ artifacts and DECISIONS.

^{Reviewed by Cursor Bugbot for commit 8dd603e. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: 9464a800-b9df-4eca-8e37-7e8661e4c81a)

[chatgpt-codex-connector]

Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-11T10:51:02.451790Z	8dd603e	Manual request

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e6264b4f92

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: 41ef3fc4-b042-4855-872b-52c8619db906)

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: 661de598-dbce-4c89-b126-5346a7af1f74)

[lx-0]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8dd603e504

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Resolve equal-timestamp default refs explicitly

When the fetched origin/<default> has advanced beyond the local default branch but both tip commits have the same second-level committer timestamp, this sort ties and selects refs/heads/<default> first; I reproduced that ordering locally, causing feature branches and worktrees to display stale progress. Git documents that committerdate is a numeric sort key and that --count=1 stops after the first ref (git-for-each-ref documentation), so use ancestry or an explicit tie-break that favors the unequal remote ref rather than treating the timestamp as a total ordering.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Guard the optional segment in set -e scripts

When a user's status-line script enables set -e, running this documented snippet outside a ytstack project, with a malformed progress block, or before the plugin is installed terminates the entire status-line script at this assignment because the substitution returns 1. Bash gives an assignment-only command the status of its command substitution (Bash reference), so handle the expected nonzero status here (for example with || true) before printing the rest of the status line.

Useful? React with 👍 / 👎.